Molecular Genetic and Physical Analysis of Gas Vesicles in Buoyant Enterobacteria

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Apollo

Environmental Microbiology (2016) 18(4), 1264–1276 doi:10.1111/1462-2920.13203 doi:10.1111/1462-2920.13203

Molecular genetic and physical analysis of gas vesicles in buoyant enterobacteria

Yosuke Tashiro,1,2† Rita E. Monson,1† bacterial cells; the first time this has been performed Joshua P. Ramsay1,3 and George P. C. Salmond1* in either strain. 1Department of Biochemistry, University of Cambridge, Cambridge CB2 1QW, UK. Introduction 2 Applied Chemistry and Biochemical Engineering Bacteria can use multiple modes of taxis, including swim- Course, Department of Engineering, Graduate School of ming, twitching, swarming and flotation. Flotation is Integrated Science and Technology, Shizuoka enabled through the regulated production of intracellular University, Hamamatsu 432-8561, Japan. buoyancy chambers: gas vesicles. Gas vesicle-driven 3 Curtin Health Innovation Research Institute Biosciences buoyancy facilitates the movement of photosynthetic Precinct, Faculty of Health Sciences, Curtin University, cyanobacteria into water column positions, allowing expo- Bentley, WA 6102, Australia. sure to wavelengths of light that can support phototrophy (Pfeifer, 2012). Gas vesicles can also increase bacterial Summary surface area-to-volume ratios, and thereby help survival under environmental and nutritional stresses (Houwink, Different modes of bacterial taxis play important roles 1956; Walsby, 1972). in environmental adaptation, survival, colonization Gas vesicles are visible as light-refracting structures and dissemination of disease. One mode of taxis is under phase-contrast microscopy (PCM), and bacterial flotation due to the production of gas vesicles. Gas colonies producing gas vesicles can be opaque on plates. vesicles are proteinaceous intracellular organelles, All gas vesicles identified to date appear to be constructed permeable only to gas, that enable flotation in aquatic from homologous proteins (Pfeifer, 2012). The primary niches. Gene clusters for gas vesicle biosynthesis gas vesicle structural protein, GvpA, assembles into are partially conserved in various archaea, tandem arrays that form ribs of the cylindrical vesicle cyanobacteria, and some proteobacteria, such as the (Englert and Pfeifer, 1993; Walsby, 1994). A second enterobacterium, Serratia sp. ATCC 39006 (S39006). protein, GvpC, forms an exterior mesh on the gas vesicle Here we present the first systematic analysis of the surface, providing further structural support and influenc- genes required to produce gas vesicles in S39006, ing gas vesicle shape (Hayes et al., 1988; 1992; Walsby identifying how this differs from the archaeon and Hayes, 1988; Englert and Pfeifer, 1993; Walsby, Halobacterium salinarum. We define 11 proteins 1994; Offner et al., 2000). Recently, the structure of GvpF essential for gas vesicle production. Mutation of gvpN from the cyanobacterium, Microcystis aeruginosa was or gvpV produced small bicone gas vesicles, sug- solved, and through electron microscopy, GvpF was gesting that the cognate proteins are involved in the shown to form a part of the gas vesicle structure (Xu et al., morphogenetic assembly pathway from bicones to 2014). However, the precise biochemical roles of many mature cylindrical forms. Using volumetric compres- other proteins found in gas vesicle gene clusters are sion, gas vesicles were shown to comprise 17% unknown, although their stoichiometry is thought to be of S39006 cells, whereas in Escherichia coli important (Shukla and DasSarma, 2004; Chu et al., 2011; heterologously expressing the gas vesicle cluster in a Pfeifer, 2012). deregulated environment, gas vesicles can occupy Recently, we discovered gas vesicles in the enterobac- around half of cellular volume. Gas vesicle produc- terium, Serratia spp. ATCC39006 (S39006), a genetically tion in S39006 and E. coli was exploited to calculate tractable host (Ramsay et al., 2011). S39006 is a the instantaneous turgor pressure within cultured Gram-negative bacillus that is pathogenic to plant and nematode hosts and produces two secondary metabolites: the tripyrrole red pigment, 2-methyl-3- pentyl-6-methoxyprodigiosin (prodigiosin) and the Received 9 November, 2015; accepted 29 December, 2015. *For β correspondence. E-mail [email protected]; Tel. +01223 333650; Fax -lactam antibiotic, 1-carbapen-2-em-3-carboxylic acid (a +01223 766108. †These authors contributed equally to this work. carbapenem) (Coulthurst et al., 2005; Williamson et al.,

V©C 2016 The Authors. Environmental Microbiology published published by by Society Society for for Applied Applied Microbiology Microbiology and and John John Wiley Wiley & & Sons Sons Ltd. Ltd. This is an open access article under the terms of the Creative Commons Attribution License, License, which which permits permits use, use, distribution distribution and and reproduction in any medium, provided provided the the original original work work is is properly properly cited. cited. 2 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C.Characterization Salmond of gas vesicles in enterobacteria 1265

2006). The genetic locus for gas vesicle biosynthesis in GvpA domain. In addition, S39006 has three genes S39006 was determined, and gas vesicles were shown to (gvpF1, gvpF2 and gvpF3) encoding homologues of a be responsible for the opaque colony phenotype and reportedly minor structural protein, GvpF, and each is buoyancy. Transcription of the gas vesicle gene cluster is conserved in γ- and δ-proteobacteria that carry a gas cell density-dependent [under the control of the SmaIR vesicle gene cluster (Fig. S1). Other genes encoding the quorum-sensing (QS) system], responsive to oxygen gas vesicle-associated proteins GvpC, GvpG, GvpH, status, and controlled by the gas vesicle regulatory GvpK and GvpN are also conserved in seven protein A, GvrA (a member of the NtrC family of regula- proteobacteria (Fig. 1 and Fig. S1). S39006 has five can- tors) encoded within the gas vesicle cluster (Ramsay didate gas vesicle proteins of unknown function: GvpV, et al., 2011; Ramsay and Salmond, 2012). Although there GvpW, GvpX, GvpY and GvpZ, and of these, GvpV and are obvious similarities between various gas vesicle gene GvpZ are conserved in four other β-, γ- and clusters, the S39006 cluster does not encode the two δ-proteobacteria gas vesicle gene clusters (Fig. 1). widely studied regulatory proteins, GvpD and GvpE, that are found in Halobacterium salinarum strains and in The gas vesicle locus in S39006 is defined by two Haloferax mediterranei (Englert et al., 1992; Ng et al., distinct operons; one directly regulated by the QS 2000; Zimmermann and Pfeifer, 2003; Hofacker et al., morphogen signal, BHL 2004). However, as in various organisms, the functionality of other proteins encoded by the gas vesicle genetic locus The 16.6 kb gas vesicle morphogenesis locus in S39006 of S39006 is unknown. is predicted to contain two operons, beginning with gvpA1 Here, we describe a comprehensive analysis of the and gvrA respectively. Reverse transcription polymerase 16.6 kb gas vesicle cluster in S39006. In frame deletions chain reaction (PCR) and 5′-RACE analysis showed that were created for each of the 19 genes in the cluster and the two operons did not overlap and that there was no impacts on gas vesicle production investigated. Pressure detectable read through between them (Fig. S2). The cor- nephelometry was used to determine the collapse pres- responding transcriptional start sites upstream of gvpA1 sure of gas vesicles, and volumetric occupancy of the and gvrA were determined (Fig. 2). Using reporter gene bacterial cytoplasm by vesicles in S39006, and an engi- fusions, we showed previously that transcription of gvpA1 neered E. coli-carrying gas vesicles, was assessed. We is controlled by the σ54, NtrC-like regulator GvrA. This is also exploited gas vesicles to calculate the instantaneous consistent with the location of a consensus σ54 binding turgor pressure of both S39006 cells and the recombinant site lying upstream of the gvpA1 transcriptional start site. E. coli strain by examining the collapse pressure of gas The production of gas vesicles in S39006 is cell density vesicles under different growth conditions. This study pre- dependent, controlled by the quorum-sensing locus, sents a complete analysis of a single gas vesicle gene SmaIR (Ramsay et al., 2011). SmaI produces the diffus- cluster and further demonstrates how gas vesicles can be ible signalling molecule, N-butanoyl-L-homoserine exploited in E. coli, generating cells that are composed of lactone (BHL) and, above a concentration threshold, BHL ∼50% gas vesicles. binds to SmaR, derepressing target gene expression (Slater et al., 2003). In a smaI mutant (BHL−), gas vesicles Results are absent but production can be restored by the addition of 1 μM BHL (Ramsay et al., 2011). BHL is therefore The core gas vesicle gene composition in defined as a diffusible morphogen. To determine whether proteobacteria SmaR directly regulated expression of gas vesicles S39006 carries a 16.6 kb cluster of 19 genes responsible through either the gvrA or gvpA1 operons, their respective for gas vesicle synthesis (Ramsay and Salmond, 2012). promoters were first inserted upstream of a This gas vesicle gene composition is similar to that β-galactosidase (β-gal) reporter fusion. We investigated of other proteobacteria, particularly β-, γ- and the expression of these two promoters in E. coli carrying δ-proteobacteria, possessing a gas vesicle gene cluster either the gvrA or gvpA1 promoter in the presence of a (Fig. 1). S39006 has three genes (gvpA1, gvpA2 and plasmid encoding SmaR, or with an empty vector control. gvpA3), encoding isoforms of the primary gas vesicle In the presence or in the absence of SmaR, β-gal expres- structural protein, GvpA; this is consistent with six sion from the gvrA promoter was the same (Fig. S2). In proteobacteria that possess a gas vesicle gene cluster. A contrast, activity of the gvpA1 promoter decreased by phylogenetic tree of GvpA showed that GvpA2 and GvpA3 over 75% in the presence of SmaR, but this repression are homologous to proteins previously annotated as could be relieved by the addition of 1 μM BHL (Fig. 2). GvpS and GvpJ, respectively, in other bacteria (Fig. S1 Together, these results suggest that QS-dependent regu- and Table S1). Despite differing annotations across lation of flotation in S39006 acts via direct transcriptional organisms, all three GvpA variants contain a canonical repression of gvpA1, but not gvrA.

VC 2016 The© Authors.2016 The Environmental Authors. Environmental Microbiology Microbiology published by published Society byfor Society Applied for Microbiology Applied Microbiology and John Wiley and John & Sons Wiley Ltd., & Sons Ltd, Environmental Microbiology, 18, 1264–1276 Environmental Microbiology 1266 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C. SalmondCharacterization of gas vesicles in enterobacteria 3

Fig. 1. Organization of gas vesicle gene clusters. Gas vesicle gene clusters from the indicated organisms were annotated and compared. Genes of predicted similar function are denoted in the same colour. Genes predicted to encode gas vesicle regulatory proteins (Gvr proteins) are indicated as hollow arrows and gas vesicle proteins (Gvp proteins) are indicated as ﬁlled arrows. The scale bar indicates 1 kb.

Fig. 2. Operonic determination of S39006 gas vesicle cluster. A. Genetic organization of the S39006 gas vesicle gene cluster locus. B. The gvpA1 transcriptional start site determined by 5′RACE is denoted in bold with a +1 above, and potential −12 and −24 sites of its promoter are underlined. Consensus sequences for σ54 are shown below. The translational start site is indicated with an arrow. C. The gvrA transcriptional start site, determined by 5′RACE, is denoted in bold with a +1 above, and potential −10 and −30 sites of its promoter are underlined. The consensus sequences for σ70 are shown below and the translational start site is indicated with an arrow.

D. SmaR regulates gvpA1 directly, but not gvrA. Escherichia coli strains carrying pRW50-gvpA1pro were grown in the presence of either pQE80 or pQE80-SmaR with added dimethylsulphoxide (DMSO) or 1 μM BHL dissolved in DMSO, where indicated. After 8 h of growth, samples were −1 taken and β-gal activity assayed (represented as RFU OD600 ). The values are the average of three biological replicates ± standard deviation (SD).

Comprehensive deletion analysis of the gas vesicle for cell buoyancy (Fig. 3); none of the corresponding operons mutant strains produced gas vesicles detectable by TEM or by PCM under any growth conditions tested. In a gvrC Previous studies of the proteins required for construction of mutant, a few gas vesicles were observed in some cells gas vesicles in archaea and cyanobacteria have focused taken from a colony on a plate, but not in planktonic cells. largely on the main structural protein, GvpA, and the Ectopic synthesis of the corresponding wild-type proteins strengthening protein, GvpC (Walsby and Hayes, 1988; restored gas vesicle production in each deletant strain, Walsby, 1994; Offner et al., 1996). In an earlier study on confirming that each in-frame knockout only affected one S39006, transposon insertion mutants and strains with gene (Fig. S3). From these data, we concluded that the reporter fusions in gvpA1 and gvrA were analysed, but protein products of these 10 genes are each required for these insertion mutations were polar (Ramsay et al., 2011) gas vesicle formation in S39006. so it was not possible to assess the impacts of the 19 S39006 mutants defective in gvpN and gvpV did not individual gvp or gvr genes (mostly with no known function) float but contained small bicone-shaped gas vesicles on gas vesicle morphogenesis. Furthermore, as our visible by TEM (Fig. 4). It has been observed in other bioinformatic analysis suggests, the core gene set found bacteria that gas vesicles initially assemble in a bicone within gas vesicle clusters is likely to be widely conserved form before morphogenetic development into the full (Fig. 1). To elucidate the role of all 19 genes in the S39006 cylindrical form (for comprehensive reviews, see Walsby, locus and to try to define the ‘minimum gene set’ required 1994; Pfeifer, 2012). When wild-type versions of either for gas vesicle morphogenesis, in frame deletions were gvpN or gvpV were expressed in these S39006 mutants, constructed for each gene. Each mutant was scored on in trans, gas vesicles were observed by PCM, and full- plates for the opaque colony morphotype (a facile indicator size mature gas vesicles were seen by TEM (Fig. S4A). of gas vesicle presence); cultures were examined directly However, neither a deletion of gvpN nor gvpV could be for phase-bright vacuoles by PCM; and flotation assays complemented by excess of the other, suggesting that the were performed to determine if mutants were buoyant two proteins perform independent functions (data not (Fig. 3). Transmission electron microscopy (TEM) images shown). Previous work demonstrated that applying pres- were also taken to observe the presence and shapes of sure to wild-type cells could lead to generation of similar gas vesicles (Fig. 4). From these data, we concluded that bicone vesicles in S39006 and, in the absence of further gvpA1, F1, G, A2, K, A3, F2 and F3 and gvrA and B were stresses, these grew into the larger form (Ramsay et al., required for the formation of phase-bright structures and 2011). However, when we applied pressure to either a

VC 2016 The© Authors.2016 The Environmental Authors. Environmental Microbiology Microbiology published by published Society byfor Society Applied for Microbiology Applied Microbiology and John Wiley and John & Sons Wiley Ltd., & Sons Ltd, Environmental Microbiology, 18, 1264–1276 Environmental Microbiology 1268 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C.Characterization Salmond of gas vesicles in enterobacteria 5

Fig. 3. Comprehensive mutational analysis of the gas vesicle production locus. Gas vesicle production in the WT and each gvp or gvr mutant was assessed throughout both the gvpA1 operon (top section) and gvrA operon (bottom section). Cells were grown in LB in sealed universals or on LB agar at 30°C for 24 h. Gas vesicle formation was confirmed by PCM observations from liquid culture (top row), flotation assays from liquid culture (second row), PCM from colonies grown on plates (third row) or colony morphology (fourth row). The scale bars in the PCM images represent 1 μm. gvpN or gvpV mutant, bicone vesicles were still visible by tional strength (Walsby and Hayes, 1988; Hayes et al., TEM, suggesting that these bicone structures are robust 1992; Pfeifer, 2012). Gas vesicles are phase bright within (Fig. S4). The measurements of gas vesicles within gvpN cells, but when collapsed, they no longer refract light. By or gvpV mutants showed that they were the same size in applying incremental pressure and examining the change both strains, although significantly different from gas vesi- in light refraction, the mean critical pressure (pc), when cles seen in the wild-type strain (Fig. S4B). Taken 50% of gas vesicle has collapsed, could be determined. together, these data suggest that GvpN and GvpV are The pc of gas vesicles in S39006 was 0.4334 MPa and, in essential for the morphogenetic assembly of gas vesicles a gvpC mutant, it decreased to 0.1435 MPa (Fig. 5A). from simple bicones into mature, functional gas vesicles This decrease in pc was counteracted by expression of in an S39006 developmental program. gvpC in trans (Fig. S5A). It was formally possible that this There are three GvpA isoforms in S39006. Ramsay et change in collapse pressure could have been because al postulated that each of the GvpA proteins might relate the gas vesicles in a gvpC mutant were a different shape to a different size of vesicle (2011). However, we found and size. However, no significant differences in gas that loss of any one of the gvpA protein isoforms pre- vesicle sizes were observed between a gvpC mutant and vented gas vesicle formation (Fig. 3). Thus, we believe wild type (Fig. S4) and so these data confirm that that all three GvpA isoforms seem to be essential for the GvpC39006 strengthens gas vesicles. production of mature gas vesicles in S39006. Gas vesicles can be produced in E. coli strain W3110 carrying the S39006 gas vesicle cluster expressed from the cosmid, pGAS (Ramsay et al., 2011). We used this Robustness of gas vesicles under pressure strain to compare the behaviour of gas vesicles in E. coli

In other organisms, GvpC is thought to form a meshed and S39006. The pc of gas vesicles in E. coli strain W3110 structure on the surface of gas vesicles, providing addi- carrying pGAS was very similar to that of the gas vesicles

Fig. 4. Transmission electron microscopy images of wild type (WT) and several gvp mutants. WT cells and mutants were grown in LB in sealed universals for 24 h and gas vesicles were observed by TEM (scale bars represent 500 nm). in S39006 (0.4012 MPa in E. coli and 0.4334 MPa in (Holland and Walsby, 2009). By measuring the relative

S39006), suggesting structural similarity in both natural change in pc when cells are exposed to medium contain- and engineered hosts (Fig. 5B). ing 0.35 M sucrose, altering the balance of solute concentration relative to the local environment, turgor pressure

(pt) in live cells can be assessed. However, until now, this Gas vesicles as tools for measurement of instantaneous has only been used in cyanobacteria and never in turgor pressure of E. coli and S39006 proteobacteria such as E. coli or S39006 (Holland and Cells maintain a cytoplasmic solute balance relative to Walsby, 2009). their environment, and the difference between external Based on TEM observations and collapse pressure and internal concentrations results in turgor pressure. proﬁles, gas vesicles in E. coli and S39006 were indis- Instantaneous measurements of turgor within cells have tinguishable. Thus, we were able to exploit them to cal- proved problematic, but a method for measuring turgor culate corresponding instantaneous turgor pressures by pressure within live cells containing gas vesicles exists pressure nephelometry. We ﬁrst examined the change in

Fig. 5. Physical characterization of gas vesicles and volumetric calculations in E. coli and S39006. (A) Collapse pressure experiments on S39006 or (B) E. coli carrying the pGAS cosmid. At 0.05 MPa pressure increments, the relative proportion of gas vesicles present in each sample was ascertained. The values are the average of three biological replicates ± SD. The black dashed lines indicate the mean collapse pressure, pc, for each sample and the continuous black line indicates pt, the turgor pressure. Where indicated, the collapse pressure was measured in cells exposed to LB medium containing 0.35 M sucrose. (C) Volumetric calculation of the percentage of gas vesicles in cells. Values indicated are the mean gas vesicle percentage of three biological replicates ± SD.

VC 2016 The© 2016Authors. The Environmental Authors. Environmental Microbiology Microbiology published by published Society by for Society Applied for Microbiology Applied Microbiology and John Wiley and John & Sons Wiley Ltd., & Sons Ltd, Environmental Microbiology, 18, 1264–1276 Environmental Microbiology 1270 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C. SalmondCharacterization of gas vesicles in enterobacteria 7 pc in S39006 between cells in Lysogeny Broth – Lennox the actinomycete, Streptomyces coelicolor contains a (LB) or cells exposed to LB + 0.35 M sucrose. The dif- cryptic gas vesicle gene cluster, but no conditions have ference in collapse pressures between these two condi- yet been identified where gas vesicles are formed, imply- tions (pt) was 0.1036 MPa (Fig. 5A), suggesting that the ing that gas vesicles may perform another, unidentified, turgor pressure in S39006 was around 0.1 MPa. Similar function in this soil bacterium (van Keulen et al., 2005). An experiments testing the difference in pc in E. coli indi- alternative explanation may be that the organism simply cated that pt was 0.1423 MPa (Fig. 5B). Thus, we have lacks one or more proteins that are essential for gas been able to exploit the presence of gas vesicles to vesicle production. The gas vesicle gene cluster in determine turgor pressure for the first time in S. coelicolor lacks gvpN, gvpV, gvpF2, gvrA, gvrB and enterobacteria such as E. coli. gvrC, all genes required for full cylindrical gas vesicle formation in S39006. Further, the GvpA protein encoded in the S. coelicolor cluster differs significantly from other Transcription of gvpA1 is independently regulated by gas vesicle-producing organisms studied so far (Fig. S1 GvrA, GvrB and GvrC and van Keulen et al., 2005). Although there is variation We showed previously that transcription of gvpA1 was between the different gene clusters, the required gene set dependent on prior expression of the gvrA-gvrC operon, described here may provide a starting point for ‘turning on’ but it was unknown which specific genes were involved in gas vesicles in S. coelicolor and other bacterial species. this regulation (Ramsay et al., 2011) To study this further, Although not all gas vesicle gene clusters contain the we combined in frame mutations in gvrA, gvrB or gvrC in same genes in an identical order, they follow a general S39006 with a gvpA1::uidA reporter fusion and analysed pattern, at least within the Proteobacteria. The gene β-glucuronidase (β-gluc) activity in these strains through- cluster is divided into two operons, one starting with out growth. It was reported previously that β-gluc activity gvpA1, as in S39006. The second operon almost always increased exponentially in stationary phase in a strain begins with gvrA and includes genes with known regula- carrying a gvpA1::uidA fusion alone. However, a signifi- tory function such as gvrB and gvrC. However, in cant decrease in β-gluc activity, reflecting gvpA1 expres- Psychromonas ingrahamii 37 and Desulfomonile tiedjei sion, was observed in strains carrying gvrA, gvrB or gvrC DCB-1 the order of the operons has been inverted, and in mutations (Fig. S6). Taken together, these results suggest Burkholderia thailandensis MSMB43 the two operons are that GvrA, GvrB and GvrC are required independently for divergently transcribed. Despite these clear differences, full expression of the gvpA1 operon and gas vesicle syn- the arrangement of genes is largely similar across many thesis in S39006. bacterial species (Fig. 1). Gas vesicles are produced by the enterobacterium, S39006, a highly genetically tractable strain (Ramsay Discussion et al., 2011). Furthermore, the gas vesicle genetic module Different taxis systems enable bacterial dissemination from 39006 can be functionally reconstituted in E. coli, between niches. The components that make up the appa- and so the roles of each S39006 gene involved in gas ratus required for swimming and twitching motility have vesicle production in both natural and heterologous hosts been widely studied and their assembly pathways are well can also be addressed. Here, we have performed the first defined (Macnab, 2003; Busch and Waksman, 2012). full mutational analysis of the two operons that form the Although the importance of gas vesicles in facilitating gas vesicle cluster in S39006 and determined the minimal flotation is understood, many organisms that produce gas gene set required for gas vesicle synthesis. We have vesicles, such as the archaeon H. salinarum, contain gas ascertained further information about proteins such as vesicle genetic clusters that contain genes absent from GvpN and GvpV, potentially acting as chaperones, that S39006, for example gvpI and gvpO, and a different regu- facilitate formation of cylindrical vesicles from small latory hierarchy, most notably well-characterized regula- bicones in S39006. In H. salinarum PHH1, the deletion of tors GvpD and GvpE (Kruger et al., 1998; Hofacker et al., gvpN resulted in smaller bicone gas vesicles, but this 2004; Scheuch et al., 2008; Bleiholder et al., 2012; study was unable to define the function of GvpN and the Marschaus and Pfeifer, 2012). Therefore, comprehensive corresponding mutations could not be complemented studies of gas vesicles in H. salinarum or the (Offner et al., 2000). Furthermore, in H. salinarum NRC-1, cyanobacterium Anabaena flos-aquae are not necessarily deletion of gvpN from the mini-chromosome pNRC100 readily comparable with those in S39006 (Hayes and also led to the formation of smaller vesicles (DasSarma Powell, 1995; Offner et al., 1996; 1998; 2000; Li and et al., 1994). However, neither of these gas vesicle gene Cannon, 1998; Hofacker et al., 2004). Furthermore, not all clusters contains a copy of gvpV, similar to that found in bacteria that possess a putative gas vesicle genetic S39006. GvpV homologues have been identified in only cluster may actually produce gas vesicles. For example, five proteobacterial species, suggesting that this protein

© 2016 TheVC Authors.2016 The Environmental Authors. Environmental Microbiology Microbiology published published by Society by for Society Applied for Microbiology Applied Microbiology and John and Wiley John & Sons Wiley Ltd, & Sons Ltd., Environmental Microbiology Environmental Microbiology, 18, 1264–1276 8 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C.Characterization Salmond of gas vesicles in enterobacteria 1271 may perform a function specific to these bacteria (Fig. 1). 1972). Furthermore, as in Anabaena and other systems, However, we have shown that both GvpN and GvpV act removal of GvpC caused more than a threefold decrease independently to extend the gas vesicle shape from in the critical collapse pressure (Buchholz et al., 1993). bicone to the full cylindrical form. Although this study has Thus, it appears that the structural robustness function of clarified the function of some gas vesicle genes, there are GvpC is conserved in S39006, Anabaena and Microcystis also differences in functions between organisms. For sp. 8401. Similarly, the role of GvpA1 (or GvpA in other example, gvpH is found in all six proteobacterial strains systems) appears to be the main structural protein. we analysed, but it is absent from A. flos-aquae and Bacil- However, the role of GvpA2/A3 (or the corresponding lus megaterium (Fig. 1). In the archaeon H. salinarum proteins GvpJ/S in other systems) is still unclear. In PHH1, gvpH is present and, as in S39006, its deletion is S39006, we have annotated these three proteins as vari- not detrimental to gas vesicle formation, yet in a gvpH ants of GvpA because all three isoforms contain clear deletion of H. salinarum, gas vesicles were less stable GvpA domains, despite the different annotations used in when isolated and could not be visualized by TEM (Offner other organisms. Finally, and most importantly, we have et al., 2000). However, an earlier study demonstrated that demonstrated that all three of these GvpA variants are an insertion in gvpH resulted in significantly reduced pro- required for the morphogenesis of functional gas vesicles duction of gas vesicles from the gene cluster on the in S39006. mini-chromosome pNRC100 in H. salinarum NRC-1 An ongoing challenge in the study of cellular turgor (DasSarma et al., 1994). In contrast, in S39006, we found pressure has been the difficulty of ascertaining instanta- no change in gas vesicle shape or stability in a strain neous turgor in a living culture without damaging the cells. carrying a gvpH mutation. For example, atomic force microscopy has been used to From previous TEM studies, it appeared that differently measure the deformability of the cellular membrane sized vesicles were made by S39006 (Ramsay et al., (Arnoldi et al., 2000; Deng et al., 2011). Although this 2011), but an extensive microscopy analysis in this study technique is useful, it can only be done at a single-cell has not shown statistically robust evidence supporting this level, which may not reflect the average turgor pressure interpretation. Our results could suggest a finely tuned across a bacterial culture. However, the biosynthesis of assembly process with few visually obvious intermediate gas vesicles can be exploited to determine the turgor steps, as almost all of our mutants were either gas vesicle pressure across cells in a heterogeneous culture. As gas negative or revealed gas vesicles with no discernible phe- vesicles are opaque, their collapse results in a change in notypic difference from those of the wild type. These culture turbidity. By monitoring the changes in pressure results echo those observed in other studies of gas collapse under different conditions (say one where turgor vesicle genetic clusters from H. salinarum NRC-1 and pressure has been removed by placing the cells in an PHH1 and in M. aeruginosa, where very few intermediate isotonic solution), the turgor pressure across a culture can assembly steps were identified (DasSarma et al., 1994; be calculated. As we can now reproduce this in E. coli, for Offner et al., 2000; Mlouka et al., 2004). the first time, the effect of particular mutations in solute It seems unlikely that the deletion of these ‘non- transport pathways, on internal cellular pressure can be essential’ genes has no impact whatsoever in gas vesicle determined. Potentially, this has widespread applications biogenesis. For example, in these experiments, we cannot and may have utility in other bacterial systems, wherever assess the kinetics of gas vesicle assembly because we they are capable of producing gas vesicles. only monitor the behaviour of cell populations. Thus, it is This study significantly enhances our understanding of formally possible that some of these ‘non essential’ genes gas vesicles in S39006, and raises multiple questions. might play very subtle roles in modulating gas vesicle First, although some roles for GvpA and GvpC have been formation that cannot be detected in crude assays or seen documented in other systems, important questions by TEM. It is also possible that S39006 modulates the remain about the functionality of proteins GvpA2 and A3 structure of gas vesicles under varying physiological con- (or GvpJ and S in other systems). All three isoforms of ditions that were not investigated in this work. GvpA (A1, A2 and A3) in S39006 contain a putative ‘GvpA’ In addition to defining the minimum gene set required domain, and our mutational and complementation analy- for gas vesicle formation in S39006, we were also able to ses have now proved that all three GvpA variants are physically characterize these gas vesicles. The collapse essential for gas vesicle formation in 39006. More infor- pressure of gas vesicles in S39006 (0.4334 MPa) is mation is clearly required about the structures of all three similar to that found in the cyanobacterium, Microcystis GvpA variants, how they may perhaps dock together, sp. 8401 (0.468 MPa) (Holland and Walsby, 2009) but is and their precise morphogenetic roles in gas vesicle very different than the collapse pressure of gas vesicles in development. Microcystis sp. BC 84/1 (0.77 MPa; Walsby and This work with in frame mutants has also defined Bleything, 1988) or H. salinarum (0.09 MPa; Walsby, several new proteins as essential for gas vesicle

VC 2016 The© 2016Authors. The Environmental Authors. Environmental Microbiology Microbiology published by published Society by for Society Applied for Microbiology Applied Microbiology and John Wiley and John & Sons Wiley Ltd., & Sons Ltd, Environmental Microbiology, 18, 1264–1276 Environmental Microbiology 1272 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C.Characterization Salmond of gas vesicles in enterobacteria 9 formation in S39006. For example, what are the precise Bacterial strains and plasmids used in this study are listed in roles of GvrA, B and C in the regulation of gas vesicle Table S2. Additional details of construction of strains and synthesis? Each of the three proteins is essential for gas plasmids are found in the Strain construction and plasmid cloning section below. Oligonucleotides used are listed in vesicle formation under the growth conditions tested here, Table S3. S39006 and E. coli were grown in LB (10 g and each independently regulates transcription of the tryptone l−1, 5 g NaCl l−1, 5 g yeast extract l−1) or LB agar gvpA1 operon, although they could be integrating different (1.5%) at 30°C or 37°C, respectively, and supplemented with environmental or physiological cues in the signal antibiotics where necessary. transduction pathway to gas vesicle development. The specific role of the three GvpF variants in gas vesicle Strain construction and plasmid cloning assembly also remains unclear. GvpF is widely conserved across different gas vesicle gene clusters (Fig. 1), and the In frame mutants carrying deletions of each gene in the protein has been identified in two proteomic gas vesicle S39006 gas vesicle cluster were constructed by homologous analyses (Shukla and DasSarma, 2004; Chu et al., 2010). recombination using derivatives of the suicide vector pKNG101 listed in Table S2. For construction of gvpA1 A recent report described the structure of GvpF from mutant, fragments flanking the open reading frame were M. aeruginosa and showed that it localized to the gas- amplified with oligonucleotide pairs, gvpA1_5F_BamHI/ facing surface of gas vesicles, suggesting that this protein gvpA1_5R_XhoI, and gvpA1_3F_XhoI/gvpA1_3R_SpeI (all does play a structural role (Xu et al., 2014). However, the oligonucleotides are listed in Table S3). These fragments precise role of GvpF in the structure or assembly of gas were used to perform overlap extension PCR, creating a DNA vesicles has never been elucidated. It also remains fragment lacking the gvpA1 open reading frame, but contain- unclear why S39006 contains three GvpF isoforms. An ing the flanking DNA. The resulting DNA fragment was digested with BamHI/SpeI and ligated into compatibly earlier hypothesis that each GvpF variant might pair with digested pKNG101 to yield pKNG101-ΔgvpA1. The one of the three GvpA proteins (Ramsay and Salmond, Escherichia coli β 2163 harbouring this plasmid was used as 2012) now seems unlikely, but further study of the struc- a donor for conjugation, and marker-exchange mutagenesis tures and functions of these essential GvpF variants is was carried out. The confirmation of the gene deletion was by required to clarify their role(s). PCR and sequence analysis. The construction of other In summary, although the production of gas vesicles mutants was performed in the same way with restriction was originally described in 1895, the specific role of each enzymes corresponding to their respective oligonucleotides. For the construction of pQE80oriT-gvpA1, the gvpA1 gene protein required for gas vesicle formation is still being was amplified from the S39006 chromosome with oligonu- dissected (DasSarma et al., 1994; Offner et al., 2000; cleotides gvpA1F-EcoRI and gvpA1R-HindIII (Table S3), Mlouka et al., 2004; Pfeifer, 2012; Tavlaridou et al., 2013; digested with EcoRI/HindIII and cloned in pQE80oriT under 2014; Xu et al., 2014). This work describes the minimal control of the lac promoter. The plasmid-producing GvpA1 gene set required for gas vesicles in S39006 and demon- without a His6 tag was introduced into Serratia by conjugation β strates that gas vesicles formed in E. coli and S39006 are using E. coli 2163 as a donor strain. The construction of physically indistinguishable. Gas vesicles have potential other plasmids containing gvp or gvr gene was performed similarly but with restriction enzymes corresponding to their applications in multiple fields, including in nanoparticles respective oligonucleotides. (DasSarma et al., 2014), as a contrast agent in ultrasound The promoter fusions of gvrApro and gvpA1pro were created imaging (Shapiro et al., 2014a,b), and for delivery of anti- by PCR using oligonucleotides oREM399/400 and gens as fusion proteins (DasSarma et al., 2014). In the oREM397/398 respectively. These fragments were digested synthetic biology, chemical factory, biotechnology and with EcoRI/HindIII and ligated into compatibly digested α bioprocessing arenas, the gas vesicle cluster could be pRW50 and used to transform E. coli strain DH5 , creating exploited in E. coli, or other organisms, to engineer cells pRW50-gvpA1pro and pRW50-gvrApro. Each of these strains was transformed with either pQE80 or pQE80-SmaR (Slater that float or sink in a tightly controlled way – perhaps in et al., 2003). response to specific synthetic chemical signals. Further- more, through the use of pressure nephelometry, the exploitation of the gas vesicle cluster in E. coli can now RNA studies allow quantifiable assessments of the impact of any muta- The cells from cultures grown in LB medium were pelleted by tion or environmental condition on the instantaneous centrifugation, and RNA was extracted from the samples turgor pressure of a whole culture. using the RNeasy mini kit (Qiagen) according to the manufacturer’s instruction. Residual DNA was removed using the TURBO DNA-free kit (Ambion) according to the manufactur- Experimental procedures er’s instructions. The transcriptional start sites of gvpA1 and ′ ′ Bacterial strains, plasmids and culture conditions gvrA were determined using a 5 /3 RACE kit (Roche). Reverse transcription PCR was carried out to determine if the Serratia sp. ATCC39006 LacA was the wild-type parental left gene cluster is operonic. The synthesis of cDNA was strain for all mutant strains constructed and used in this work. performed with random hexamers using SuperScript II (Life

Technologies). PCR was conducted using cDNA as a tem- apparatus shown in Fig. S5. As intracellular gas vesicles plate with oligonucleotide pairs: RTGVL1/RTGVL2, RTGVL3/ collapsed under increasing pressure, changes in culture tur- RTGVL4, RTGVL5/RTGVL6, RTGVL7/RTGVL8, RTGVL9/ bidity were assessed. The collapse pressures of gas vesicles RTGVL10, RTGVL11/RTGVL12, RTGVL13/RTGVL14, were determined by pressure nephelometry. The machine RTGVL15/RTGVL16 and RTGVL17/RTGVL18. was set to zero with 4 ml of LB and 500 μl of the indicated The transcriptional start site of gvpA1 was determined by culture was added and mixed. This mixture was carefully 5′RACE using a 5′/3′ RACE kit (Roche). The synthesis of placed into the nephelometer, ensuring that 100% of gas cDNA was carried out with an oligonucleotide RTgvpCR and vesicles remained intact, and the millivoltmeter set to 100. cDNA was treated with terminal transferase and dATP as Pressure was applied as N2 gas in 0.05 MPa increments, up per the manufacturer’s instruction. PCR for 5′RACE was to 1 MPa, and the change in culture turbidity monitored after carried out using specific oligonucleotides complementary 20 s of equilibration using an analogue millivoltmeter. At each for the gvpA1 (gvpA1R-HindIII) and oligo-dT. The PCR pressure increment, the proportion of gas vesicle remaining product was cloned into pBLUEScript SK+, yielding (Gp) was calculated using the following formula: pBluescript + gvpA1-5′RACE. Both the original and cloned GG− 5′RACE products were sequenced. G = m 1MPa p − 5′RACE analysis was also used to identify the transcrip- GG01MPa MPa tional start site of gvrA and to determine if the right gene where Gm is the measurement taken at the indicated pressure cluster was operonic. cDNA synthesis was carried out using and G and G are the measurements of turbity taken at ′ 1MPa 0MPa an oligonucleotide complementary to the 3 end of gvrC, the 1 MPa and 0 MPa respectively.The mean critical pressure final gene in the predicted operon containing gvrA, (p ) was calculated by determining the pressure when gvpH, gvpZ, gvpF2, gvpF3, gvrB and gvrC. cDNA was c Gp = 0.5. When cultures were exposed to 0.35 M sucrose treated with terminal transferase and dGTP to attach a poly-G solutions, the mean apparent critical pressure (p ) was also tract to the 3′ of the cDNA. PCR for 5′RACE was carried out a determined by identifying the pressure when Gp = 0.5. The using specific oligonucleotides complementary to gvrA turgor pressure (p ) was calculated by subtracting p from p . (gvrASP1bam) and a poly-C oligonucleotide (polyDRACE). t a c Nested oligonucleotides were then used to further amplify from this reaction (gvrASP2bam and Anchorprimer, see Volume calculation of gas vesicles Roche instructions), which resulted in amplification of a single product that was subsequently cloned into pBLUEScript SK+. Calculation of gas volume in cultures was performed as Both the original and the cloned 5′RACE products were described in Walsby (1982). Briefly, cultures were grown as sequenced. described for pressure nephelometry, and the relative change in volume was calculated using the apparatus shown in Fig. S5. The change in culture volume was recorded under Confirmation of gas vesicle formation pressure that caused the collapse of 100% of vesicles (0.8 MPa). The change in culture volume was normalized to A total of 5 ml of overnight cultures was grown in 25 ml sealed the total cell numbers to determine the percentage of cell universals on a roller wheel at 30°C and then used to sub- volume that was due to gas vesicles. Volumetric calculations culture into a second 5 ml broth, which was grown for 24 h were performed using a compression tube (Walsby, 1982; under the same conditions. For flotation assays, cultures Walsby et al., 1992). Tubes, with a capillary 0.2 mm in inter- were then kept statically at room temperature for 48 h before nal diameter, were filled to capacity (∼2.5 ml of culture). The imaging by PCM, TEM or flotation in the universal tube was volume of gas vesicles was determined by observing the assessed. Cells from liquid cultures or agar plates were relative change in the meniscus inside the capillary (L) upon observed by PCM or TEM as described previously (Ramsay application of 0.8 MPa of pressure, via N gas, causing the et al., 2011). Cells were imaged for PCM using an Olympus 2 collapse of all gas vesicles within the culture. A change in BX-51 microscope with a 100x oil-immersion lens, with a 1 mm of the meniscus within a capillary 0.2 mm in diameter is QICAM monochrome camera and QCAPTURE PRO-6 soft- equivalent to 31.4 ηl. The mean gas vesicle volume per cell ware. For TEM, samples were observed in a FEI Tecnai G2 was calculated using the change in volume normalized to the TEM. Images were captured with an AMT XR60B digital number of cells in each culture, determined by colony counts. camera running DEBEN software at the Cambridge University Advanced Imaging Centre. For colony morphology experiments, 3 µl of normalized culture were spotted onto LB agar Measurements of gas vesicles plates, incubated at 30°C for 24 h and colony morphologies observed. Cells scraped from plates were suspended in PBS TEM images were analysed using IMAGEJ (Schneider et al., and observed by PCM. 2012). Each image was calibrated against its scale bar, and gas vesicles were measured. Mean heights and widths of gas vesicles were then calculated. Pressure nephelometry

Pressure collapse experiments were performed largely using β-Galactosidase and β-glucuronidase activity the same apparatus, and as previously described, in Holland and Walsby (2009). Cultures were grown overnight on a roller S39006 Cultures were grown at 30°C with shaking at wheel at 30°C and then left to settle for 24 h at room tem- 250 r.p.m. and the optical densities (OD600) examined every perature before application of incremental pressure using the 2 h. After 6 h of incubation, samples were taken and frozen at

−80°C until required. Activity of β-gluc, the protein product of Buchholz, B.E., Hayes, P.K., and Walsby, A.E. (1993) The uidA, was determined using the fluorogenic substrate, distribution of the outer gas vesicle protein, GvpC, on the 4′-Methylumbelliferyl–d-glucuronide (Melford Laboratories), Anabaena gas vesicle, and its ratio to GvpA. J Gen as described previously (Ramsay et al., 2011). Briefly, frozen Microbiol 139: 2353–2363. samples were thawed and diluted to appropriate ratios where Busch, A., and Waksman, G. (2012) Chaperone-usher path- a linear increase in activity could be observed, and 10 µlof ways: diversity and pilus assembly mechanism. Philos diluted samples was frozen and thawed again. Then, 100 µl Trans R Soc Lond B Biol Sci 367: 1112–1122. of reaction buffer (400 g ml−1 lysozyme, 250 g ml−1 Chu, L.J., Chen, M.C., Setter, J., Tsai, Y.S., Yang, H., Fang, 4′-Methylumbelliferyl–d-glucuronide in PBS) was added, and X., et al. (2010) New structural proteins of Halobacterium fluorescence was immediately monitored (excitation 360 nm, salinarum gas vesicle revealed by comparative proteomics emission 450 nm, cut-off 435 nm) every 30 s for 30 min using analysis. J Proteome Res 10: 1170–1178. a Gemini XPS plate reader. β-Gluc expression was measured Chu, L.J., Chen, M.C., Setter, J., Tsai, Y.S., Yang, H., Fang, by the relative fluorescence units per second, produced from X., et al. (2011) New structural proteins of Halobacterium cleavage of 4′-Methylumbelliferyl-β-d-glucuronide, normal- salinarum gas vesicle revealed by comparative proteomics −1 −1 ized to the optical density of the culture (RFU s OD600 ). analysis. J Proteome Res 10: 1170–1178. β-Gal activity was monitored in a similar fashion but using Coulthurst, S.J., Barnard, A.M., and Salmond, G.P. (2005) 4′-Methylumbelliferyl-β-d-galactopyranoside (Melford Labo- Regulation and biosynthesis of carbapenem antibiotics in ratories) as a fluorogenic substrate. bacteria. Nat Rev Microbiol 3: 295–306. DasSarma, P., Negi, V.D., Balakrishnan, A., Karan, R., Bioinformatic analyses Barnes, S., Ekulona, F., et al. (2014) Haloarchaeal gas vesicle nanoparticles displaying Salmonella SopB antigen Sequences were accessed from NCBI, and putative gas reduce bacterial burden when administered with live vesicle clusters in other organisms were identified using PSI- attenuated bacteria. Vaccine 32: 4543–4549. BLAST (Altschul et al., 1997). Amino acid sequences were DasSarma, S., Arora, P., Lin, F., Molinari, E., and Yin, L.R. aligned using CLUSTALX (Larkin et al., 2007) and phylogenetic (1994) Wild-type gas vesicle formation requires at least ten trees were constructed by NJPLOT (Perriere and Gouy, 1996). genes in the gvp gene cluster of Halobacterium halobium Phylogenetic distances were determined by neighbour- plasmid pNRC100. J Bacteriol 176: 7646–7652. joining analysis. The numbers on branching points are per- Deng, Y., Sun, M., and Shaevitz, J.W. (2011) Direct meas- centages of bootstrap values with 1000 replicates. urement of cell wall stress stiffening and turgor pressure in live bacterial cells. Phys Rev Lett 107: 158101. Acknowledgements Englert, C., and Pfeifer, F. (1993) Analysis of gas vesicle gene expression in Haloferax mediterranei reveals that The authors would like to thank Professor Tony Walsby GvpA and GvpC are both gas vesicle structural proteins. (Emeritus, Bristol University) for advice, technical help and J Biol Chem 268: 9329–9336. donation of the pressure nephelometry and volumetric calcu- Englert, C., Kruger, K., Offner, S., and Pfeifer, F. (1992) lation apparatus. We would also like to thank Alison Drew for Three different but related gene clusters encoding gas technical support, and Chin Mei Lee and Andrew Day for vesicles in halophilic archaea. J Mol Biol 227: 586– critical reading. REM and GPCS were supported through the 592. BBSRC (Grant ID BB/K001833/1). YT was supported by a Hayes, P., and Powell, R. (1995) The gvpA/C cluster of Scientific Research Fellowship from the Japan Society for the Anabaena flos-aquae has multiple copies of a gene encod- Promotion of Sciences (JSPS) and JPR was supported by a ing GvpA. Arch Microbiol 164: 50–57. Herschel Smith Post Doctoral Fellowship while at Cambridge Hayes, P.K., Lazarus, C.M., Bees, A., Walker, J.E., and University. The authors declare that they have no conflict of Walsby, A.E. (1988) The protein encoded by gvpC is a interest. minor component of gas vesicles isolated from the cyanobacteria Anabaena flos-aquae and Microcystis sp. References Mol Microbiol 2: 545–552. Hayes, P.K., Buchholz, B., and Walsby, A.E. (1992) Gas Altschul, S.F., Madden, T.L., Schaffer, A.A., Zhang, J., Zhang, vesicles are strengthened by the outer-surface protein, Z., Miller, W., and Lipman, D.J. (1997) Gapped BLAST and GvpC. Arch Microbiol 157: 229–234. PSI-BLAST: a new generation of protein database search Hofacker, A., Schmitz, K.M., Cichonczyk, A., Sartorius-Neef, programs. Nucleic Acids Res 25: 3389–3402. S., and Pfeifer, F. (2004) GvpE- and GvpD-mediated tran- Arnoldi, M., Fritz, M., Bauerlein, E., Radmacher, M., scription regulation of the p-gvp genes encoding gas vesi- Sackmann, E., and Boulbitch, A. (2000) Bacterial turgor cles in Halobacterium salinarum. Microbiology 150: 1829– pressure can be measured by atomic force microscopy. 1838. Phys Rev E Stat Phys Plasmas Fluids Relat Interdiscip Holland, D.P., and Walsby, A.E. (2009) Digital recordings of Topics 62: 1034–1044. gas-vesicle collapse used to measure turgor pressure and Bleiholder, A., Frommherz, R., Teufel, K., and Pfeifer, F. cell-water relations of cyanobacterial cells. J Microbiol (2012) Expression of multiple tfb genes in different Methods 77: 214–224. Halobacterium salinarum strains and interaction of TFB Houwink, A.L. (1956) Flagella, gas vacuoles and cell-wall with transcriptional activator GvpE. Arch Microbiol 194: structure in Halobacterium halobium; an electron micro- 269–279. scope study. J Gen Microbiol 15: 146–150.

van Keulen, G., Hopwood, D.A., Dijkhuizen, L., and Sawers, Shapiro, M.G., Goodwill, P.W., Neogy, A., Yin, M., Foster, R.G. (2005) Gas vesicles in actinomycetes: old buoys in F.S., Schaffer, D.V., and Conolly, S.M. (2014a) Biogenic novel habitats? Trends Microbiol 13: 350–354. gas nanostructures as ultrasonic molecular reporters. Nat Kruger, K., Hermann, T., Armbruster, V., and Pfeifer, F. (1998) Nanotechnol 9: 311–316. The transcriptional activator GvpE for the halobacterial gas Shapiro, M.G., Ramirez, R.M., Sperling, L.J., Sun, G., Sun, vesicle genes resembles a basic region leucine-zipper J., Pines, A., et al. (2014b) Genetically encoded reporters regulatory protein. J Mol Biol 279: 761–771. for hyperpolarized xenon magnetic resonance imaging. Larkin, M.A., Blackshields, G., Brown, N.P., Chenna, R., Nat Chem 6: 629–634. McGettigan, P.A., McWilliam, H., et al. (2007) Clustal Shukla, H.D., and DasSarma, S. (2004) Complexity of gas W and Clustal X version 2.0. Bioinformatics 23: 2947– vesicle biogenesis in Halobacterium sp. strain NRC-1: 2948. identification of five new proteins. J Bacteriol 186: 3182– Li, N., and Cannon, M.C. (1998) Gas vesicle genes identified 3186. in Bacillus megaterium and functional expression in Slater, H., Crow, M., Everson, L., and Salmond, G.P. (2003) Escherichia coli. J Bacteriol 180: 2450–2458. Phosphate availability regulates biosynthesis of two antibi- Macnab, R.M. (2003) How bacteria assemble flagella. Annu otics, prodigiosin and carbapenem, in Serratia via both Rev Microbiol 57: 77–100. quorum-sensing-dependent and -independent pathways. Marschaus, L., and Pfeifer, F. (2012) A dual promoter region Mol Microbiol 47: 303–320. with overlapping activator sequences drives the expression Tavlaridou, S., Faist, K., Weitzel, K., and Pfeifer, F. (2013) of gas vesicle protein genes in haloarchaea. Microbiology Effect of an overproduction of accessory Gvp proteins on 158: 2815–2825. gas vesicle formation in Haloferax volcanii. Extremophiles Mlouka, A., Comte, K., Castets, A.-M., Bouchier, C., and 17: 277–287. Tandeau de Marsac, N. (2004) The gas vesicle gene Tavlaridou, S., Winter, K., and Pfeifer, F. (2014) The acces- cluster from Microcystis aeruginosa and DNA rearrange- sory gas vesicle protein GvpM of haloarchaea and its ments that lead to loss of cell buoyancy. J Bacteriol 186: interaction partners during gas vesicle formation. 2355–2365. Extremophiles 18: 693–706. Ng, W.V., Kennedy, S.P., Mahairas, G.G., Berquist, B., Pan, Walsby, A.E. (1972) Structure and function of gas vacuoles. M., Shukla, H.D., et al. (2000) Genome sequence of Bacteriol Rev 36: 1–32. Halobacterium species NRC-1. Proc Natl Acad Sci USA Walsby, A.E. (1982) The elastic compressibility of gas vesi- 97: 12176–12181. cles. Proceedings of the Royal Society Series B-Biological Offner, S., Wanner, G., and Pfeifer, F. (1996) Functional Sciences 216: 355–368. studies of the gvpACNO operon of Halobacterium Walsby, A.E. (1994) Gas vesicles. Microbiol Rev 58: 94–144. salinarium reveal that the GvpC protein shapes gas vesi- Walsby, A.E., and Bleything, A. (1988) The dimensions of cles. J Bacteriol 178: 2071–2078. cyanobacterial gas vesicles in relation to their efficiency in Offner, S., Ziese, U., Wanner, G., Typke, D., and Pfeifer, F. providing buoyancy and withstanding pressure. Microbiol- (1998) Structural characteristics of halobacterial gas vesi- ogy 134: 2635–2645. cles. Microbiology 144: 1331–1342. Walsby, A.E., and Hayes, P.K. (1988) The minor Offner, S., Hofacker, A., Wanner, G., and Pfeifer, F. (2000) cyanobacterial gas vesicle protein, GvpC, is attached to Eight of fourteen gvp genes are sufficient for formation of the outer surface of the gas vesicle. J Gen Microbiol 134: gas vesicles in halophilic archaea. J Bacteriol 182: 4328– 2647–2657. 4336. Walsby, A.E., Kinsman, R., and George, K.I. (1992) The Perriere, G., and Gouy, M. (1996) WWW-query: an on-line measurement of gas vesicle volume and buoyant density in retrieval system for biological sequence banks. Biochimie planktonic bacteria. J Microbiol Methods 15: 293–309. 78: 364–369. Williamson, N.R., Fineran, P.C., Leeper, F.J., and Salmond, Pfeifer, F. (2012) Distribution, formation and regulation of gas G.P. (2006) The biosynthesis and regulation of bacterial vesicles. Nat Rev Microbiol 10: 705–715. prodiginines. Nat Rev Microbiol 4: 887–899. Ramsay, J.P., and Salmond, G.P.C. (2012) Quorum sensing- Xu, B.Y., Dai, Y.N., Zhou, K., Liu, Y.T., Sun, Q., Ren, Y.M., controlled buoyancy through gas vesicles: Intracellular et al. (2014) Structure of the gas vesicle protein GvpF from bacterial microcompartments for environmental adapta- the cyanobacterium Microcystis aeruginosa. Acta tion. Commun Integr Biol 5: 96–98. Crystallogr D Biol Crystallogr 70: 3013–3022. Ramsay, J.P., Williamson, N.R., Spring, D.R., and Salmond, Zimmermann, P., and Pfeifer, F. (2003) Regulation of the G.P.C. (2011) A quorum-sensing molecule acts as a expression of gas vesicle genes in Haloferax mediterranei: morphogen controlling gas vesicle organelle biogenesis interaction of the two regulatory proteins GvpD and GvpE. and adaptive flotation in an enterobacterium. Proc Natl Mol Microbiol 49: 783–794. Acad Sci USA 108: 14932–14937. Scheuch, S., Marschaus, L., Sartorius-Neef, S., and Pfeifer, F. (2008) Regulation of gvp genes encoding gas vesicle Supporting information proteins in halophilic Archaea. Arch Microbiol 190: 333– Additional Supporting Information may be found in the online 339. version of this article at the publisher’s web-site: Schneider, C.A., Rasband, W.S., and Eliceiri, K.W. (2012) NIH Image to ImageJ: 25 years of image analysis. Nat Fig. S1. Phylogenetic relationships of GvpA, GvpF/L, GvpC, Methods 9: 671–675. GvpG, GvpH, GvpK, GvpN, GvpV and GvpZ.

VC 2016 The© Authors. 2016 The Environmental Authors. Environmental Microbiology Microbiology published by published Society for by Applied Society Microbiology for Applied Microbiology and John Wiley and & John Sons Wiley Ltd., & Sons Ltd, Environmental Microbiology, 18, 1264–1276 Environmental Microbiology 1276 Y. Tashiro, R. E. Monson, J. P. Ramsay and G. P. C.Characterization Salmond of gas vesicles in enterobacteria 13

Fig. S2. The gas vesicle production locus of S39006 is com- Fig. S4. Measurements of GVs in different mutant strains posed of two operons. (A) The locus of the ampliﬁed region in and with application of pressure. (A) gvpV or gvpN mutant the gvp cluster. (B) RT-PCR results showing that the left strains were imaged by TEM with or without 1.0 MPa pres- cluster, starting from gvpA1, is operonic. (C) RT-PCR result sure or with either gvpV or gvpN added back in trans. Scale showing that there is no read through between the left and right bars indicate 500 nm. (B) Measurement of GVs in different operon. indicates the marker lane, with indicated sizes on the strains. TEM images of GVs from the indicated strains, grown side. P indicates a sample including reverse transcriptase, C in sealed universals, were analysed using IMAGEJ and the indicates a positive control sample containing only genomic height and widths of the indicated number (below) of vesicles DNA, and N indicates a negative control sample with no measured. The values presented are the average heights reverse transcriptase. (D) SmaR does not regulate gvrA and widths ± SD. directly. Escherichia coli strains carrying pRW50-gvrApro were Fig. S5. (A) Restoration of the collapse pressure phenotype grown in the presence of either pQE80 or pQE80-SmaR with in the gvpC mutant, by an in trans copy of gvpC. (B–C) added DMSO or 1 μM BHL dissolved in DMSO, where indi- Apparatus used for pressure nephelometry (B) or volumetric cated. After 8 h of growth, samples were taken and β-gal calculations (C). −1 activity assayed (represented as RFU OD600 ). The values are Fig. S6. Expression from the gvpA1::uidA transcriptional the average of three biological replicates ± SD. fusion throughout growth in S39006 WT, ΔgvrA, ΔgvrB, and Fig. S3. Complementation of GV formation in in frame muta- ΔgvrC backgrounds. Dashed lines show growth and solid tions. Mutations in the GV cluster that failed to produce GVs lines show gene expression levels via reporter enzymes. were grown with the indicated plasmid, with or without β-Gluc activity was assayed and represented as RFU −1 0.1 mM IPTG, and the cells were observed by PCM, ﬂotation OD600 . Values are the mean of three biological repli- assays and colony morphology. (A) Analyses of ΔgvpA1, cates ± SD. ΔgvpA2 and ΔgvpA3 mutants. (B) Analyses of ΔgvpF1, Table S1. Comparison of amino acid sequences of GvpA ΔgvpF and ΔgvpF3 mutants. (C) Analyses of ΔgvpG and and GvpF proteins from S39006. ΔgvpK mutants. (D) Analyses of ΔgvpN, ΔgvpV, ΔgvrA, ΔgvrB Table S2. Bacterial strains and plasmids used in this study. and ΔgvrC mutants. Scale bar indicates 1 μm. Table S3. Oligonucleotides used in this study.